Table 2 Estimation of cLOS with respect to model A (no censoring of deaths) as well as cLOS (discharged) and cLOS (death) (based on model A but distinguishing between death and discharge). Moreover, cLOS with respect to model B (censoring of deaths). Additionally we calculate the bias between model A and model B and the bias between model B and cLOS based on model A for discharged patients only. The comparison is done for the estimation of cLOS by assuming constant hazard and by using the etm package (assuming time-dependent hazards)